The case of the missing model The modernisation
- Slides: 69
The case of the missing model: The modernisation of multivariate analysis in ecology David Warton School of Mathematics and Statistics
The case of the missing model: The modernisation of multivariate analysis in ecology • Introduction • The Bray-Curtis distance and other 1980’s memorabilia • Building a model for multivariate abundances in ecology • A 2020 vision for multivariate analysis in ecology David Warton – Modernising multivariate analysis in ecology
An exciting time to do statistics! http: //www. allaboutapple. com Speed: 1 MHz RAM: 64 k. B Matthew Riegler, CC-BY, Wikimedia Commons Speed: 3. 8 GHz RAM: 16 GB Computing revolution → Data analysis revolution David Warton – Modernising multivariate analysis in ecology
New technology → New data Rashid et al. (2012) Evolutionary Bioinformatics Warton & Shepherd (2010) Annals of Applied Statistics David Warton – Modernising multivariate analysis in ecology Yamaguchi et al. (2011) Cancer Informatics
New technology → New analysis methods → Francis Hui (2014) Dunstan et al. (2013) Journal of Agricultural, Biological and Environmental Statistics David Warton – Modernising multivariate analysis in ecology
Example: multivariate abundances in ecology Widely used to study communities. How do grassland invertebrates respond to climate? Alex Brown (2010) David Warton – Modernising multivariate analysis in ecology 36 sites 64 Hemiptera species 0 0 0 2 0 0 3 0 0 4 0 0 0 ⁞ 4 5 1 2 0 0 0 0 0 3 0 ⁞ 1 0 0 0 0 1 1 ⁞ 0 1 0 0 0 0 1 ⁞ 0 0 1 0 0 0 0 0 12 ⁞ 0 0 0 0 2 6 1 ⁞ 1 0 0 0 1 0 ⁞ 0 0 0 0 3 0 0 0 0 ⁞ 0 0 1 0 0 0 0 ⁞ 0 … … … … 0 1 0 0 2 0 0 0 0 ⁞ 0
Multivariate abundances: two key properties • Abundance Different species, different variability (mean-variance relationship) 0 • Multivariate Lots of species (high dimensional) Big problem! David Warton – Modernising multivariate analysis in ecology 0 0 2 0 0 3 0 0 4 0 0 0 ⁞ 4 5 1 2 0 0 0 0 0 3 0 ⁞ 1 0 0 0 0 1 1 ⁞ 0 1 0 0 0 0 1 ⁞ 0 0 1 0 0 0 0 0 12 ⁞ 0 0 0 0 2 6 1 ⁞ 1 0 0 0 1 0 ⁞ 0 0 0 0 3 0 0 0 0 ⁞ 0 0 1 0 0 0 0 ⁞ 0 … … … … 0 1 0 0 2 0 0 0 0 ⁞ 0
The case of the missing model: The modernisation of multivariate analysis in ecology • Introduction • The Bray-Curtis distance and other 1980’s memorabilia • Building a model for multivariate abundances in ecology • A 2020 vision for multivariate analysis in ecology David Warton – Modernising multivariate analysis in ecology
1980’s solution… 0. 82 0. 69 0 0 0 2 0 0 3 0 0 4 0 0 0 ⁞ 4 5 1 2 0 0 0 0 0 3 0 ⁞ 1 0 0 0 0 1 1 ⁞ 0 1 0 0 0 0 1 ⁞ 0 0 1 0 0 0 0 0 12 ⁞ 0 0 0 0 2 6 1 ⁞ 1 0 0 0 1 0 ⁞ 0 0 0 0 3 0 0 0 0 ⁞ 0 0 1 0 0 0 0 ⁞ 0 … … … … 0 1 0 0 2 0 0 0 0 ⁞ 0 1. 00 1. 00 0. 68 0. 92 ⁞ 0. 60 0. 90 0. 60 Widely used – often taught in second year ecology! David Warton – Modernising multivariate analysis in ecology 1. 00 0. 60
Some other 1980’s trends… 0. 82 0. 69 0 0 0 2 0 0 3 0 0 4 0 0 0 ⁞ 4 5 1 2 0 0 0 0 0 3 0 ⁞ 1 0 0 0 0 1 1 ⁞ 0 1 0 0 0 0 1 ⁞ 0 0 1 0 0 0 0 0 12 ⁞ 0 0 0 0 2 6 1 ⁞ 1 0 0 0 1 0 ⁞ 0 0 0 0 3 0 0 0 0 ⁞ 0 0 1 0 0 0 0 ⁞ 0 … … … … 0 1 0 0 2 0 0 0 0 ⁞ 0 1. 00 1. 00 0. 68 0. 92 ⁞ 0. 60 0. 90 0. 60 1. 00 0. 60 David Warton – Modernising multivariate analysis in ecology
1980’s solution… no good in the 2010’s Poor power properties (transformation doesn’t help rarer spp) (because mean-variance relationship ignored) Simulation: a large (10 x) effect in one species only. Results: effect only detected in high-variance species. Warton et al. (2012) Methods in Ecology & Evolution David Warton – Modernising multivariate analysis in ecology
Transformation never works for rare counts David Warton – Modernising multivariate analysis in ecology
1980’s solution… no good in the 2010’s Can’t detect special mean structures (because no mean model) e. g. evidence of a change in composition? Simulation: 2 -fold change in total abundance, no change in composition Warton et al. (in prep) David Warton – Modernising multivariate analysis in ecology
1980’s solution… no good in the 2010’s Can’t detect special mean structures (because no mean model) David Warton – Modernising multivariate analysis in ecology
1980’s solution… no good in the 2010’s Same dataset, same question, different answers? (lacks diagnostic tools) e. g. coral abundance in the Thousand Islands, Indonesia David Warton – Modernising multivariate analysis in ecology (Warwick et al. 1990)
What’s missing from this picture? 0 0 0 2 0 0 3 0 0 4 0 0 0 ⁞ 4 5 1 2 0 0 0 0 0 3 0 ⁞ 1 0 0 0 0 1 1 ⁞ 0 1 0 0 0 0 1 ⁞ 0 0 1 0 0 0 0 0 12 ⁞ 0 0 0 0 2 6 1 ⁞ 1 0 0 0 1 0 ⁞ 0 0 0 0 3 0 0 0 0 ⁞ 0 0 1 0 0 0 0 ⁞ 0 … … … … 0 1 0 0 2 0 0 0 0 ⁞ 0 0. 82 0. 69 1. 00 0. 68 0. 92 ⁞ 0. 60 0. 90 0. 60 1. 00 0. 60 This prevents specification of a model for the data-generating mechanism David Warton – Modernising multivariate analysis in ecology
The case of the missing model: The modernisation of multivariate analysis in ecology • Introduction • The Bray-Curtis distance and other 1980’s memorabilia • Building a model for multivariate abundances in ecology • A 2020 vision for multivariate analysis in ecology David Warton – Modernising multivariate analysis in ecology
Some things that happened since the 1980’s… • Bayesian hierarchical models (MCMC) • Monte Carlo EM and modern numerical integration • PIT residuals as diagnostics for GLMs (Gelfand Smith 1990) (Wei & Tanner 1990) (Dunn & Smyth 1996) David Warton – Modernising multivariate analysis in ecology
Residuals are hard to define for count data David Warton – Modernising multivariate analysis in ecology
PIT residuals ~U(0, 1) when model is correct David Warton – Modernising multivariate analysis in ecology
We can use PIT residuals to check model fit David Warton – Modernising multivariate analysis in ecology
Some things that happened since the 1980’s… • • Bayesian hierarchical models (MCMC) Modern numerical integration, Monte Carlo EM PIT residuals to check model fit for counts Penalised likelihood (e. g. LASSO, ) Generalised linear latent variable models Graphical modelling Variational approximation David Warton – Modernising multivariate analysis in ecology (Gelfand Smith 1990) (Wei & Tanner 1990) (Dunn & Smyth 1996) Tibshirani 1996 (Skrondal & Rabe-Hesketh 2004) (Friedman et al. 2008) (Ormerod & Wand 2010)
Things to consider when building a model • What is the research Question? – descriptive, prediction, testing a priori hypothesis, … • What Properties do my data have? – abundance (mean-variance) – multivariate (high-dimensional) Warton et al. (2015) Plant Ecology David Warton – Modernising multivariate analysis in ecology
Data properties? Abundance (mean-variance) The question? Test (env impact) Composition Explain spp variation Predict (new sites or spp) Classify Ordination Study correlations David Warton – Modernising multivariate analysis in ecology Multivariate (high-dimensional)
Data properties? Abundance (mean-variance) Multivariate (high-dimensional) The question? Test (env impact) Composition Generalised linear model (GLM) Explain spp variation Predict (new sites or spp) Classify Ordination Study correlations David Warton – Modernising multivariate analysis in ecology for abundance yij of species j at site i as a function of site predictors xi
Data properties? Abundance (mean-variance) Multivariate (high-dimensional) The question? Test (env impact) Composition Explain spp variation Predict (new sites or spp) Classify Ordination Multivariate frameworks: • Generalised estimating equations (GEE) (Warton 2011 Biometrics) • Hierarchical models • Copulas (Warton et al 2015 Trends Ecol Evol) (Popovic & Warton in review) But high-dimensionality is the real issue (64 species → 2016 correlations!) Study correlations David Warton – Modernising multivariate analysis in ecology
Data properties? Abundance (mean-variance) Multivariate (high-dimensional) The question? Test (env impact) Composition Explain spp variation Predict (new sites or spp) Classify Ordination Study correlations David Warton – Modernising multivariate analysis in ecology Options for high dimensionality: • Ignore it (e. g. for prediction) Brown et al (2014) Methods Ecol Evol • Resample rows for valid inference Wang et al (2012) Methods Ecol Evol • Shrink covariance (e. g. toward identity) Warton (2008) JASA • Covariance modelling, e. g. – Factor analysis (latent variables) – Graphical modelling Warton et al (2015) Trends Ecol Evol
Data properties? Abundance (mean-variance) Composition Do grassland invertebrate communities change when host plants are transplanted? 64 Hemiptera species Explain spp variation Predict (new sites or spp) Classify Ordination Study correlations David Warton – Modernising multivariate analysis in ecology 36 sites The question? Test (env impact) Multivariate (high-dimensional) 0 0 0 2 0 0 3 0 0 4 0 0 0 ⁞ 4 5 1 2 0 0 0 0 0 3 0 ⁞ 1 0 0 0 0 1 1 ⁞ 0 1 0 0 0 0 1 ⁞ 0 0 1 0 0 0 0 0 12 ⁞ 0 0 0 0 2 6 1 ⁞ 1 0 0 0 1 0 ⁞ 0 0 0 0 3 0 0 0 0 ⁞ 0 0 1 0 0 0 0 ⁞ 0 … … … … 0 1 0 0 2 0 0 0 0 ⁞ 0
Data properties? Abundance (mean-variance) The question? Test (env impact) Multivariate (high-dimensional) Generalised linear model (GLM) Composition Explain spp variation Predict (new sites or spp) Classify Ordination Study correlations David Warton – Modernising multivariate analysis in ecology for abundance yij of species j at site i as a function of site predictors xi
Data properties? Abundance (mean-variance) The question? Test (env impact) Composition Explain spp variation Predict (new sites or spp) Classify Ordination Study correlations David Warton – Modernising multivariate analysis in ecology Multivariate (high-dimensional) Resample rows for testing How to resample GLMs? “PIT-trap” for residual resampling (Warton & Wang in review)
PIT-trap: David Warton – Modernising multivariate analysis in ecology Warton & Wang (in review)
Data properties? Abundance (mean-variance) The question? Test (env impact) Composition Explain spp variation Predict (new sites or spp) Classify Ordination Study correlations David Warton – Modernising multivariate analysis in ecology row-resample GLMs Multivariate (high-dimensional)
Data properties? The question? Abundance (mean-variance) Ordination David Warton – Modernising multivariate analysis in ecology Multivariate (high-dimensional)
Data properties? Abundance (mean-variance) Multivariate (high-dimensional) The question? Generalised linear model (GLM) Ordination David Warton – Modernising multivariate analysis in ecology
Data properties? Abundance (mean-variance) Multivariate (high-dimensional) The question? Generalised linear model (GLM) with latent variables zi Warton et al. (Trends Ecol Evol) Ovaskainen et al. (in press) Methods Ecol Evol Ordination Computation a bit tricky: The zi: (1) are ordination axes • MCMC (2) model correlation • variational approximation (Hui, “boral”, Ovaskainen & Blanchet “HMSC”) (Hui et al in review) David Warton – Modernising multivariate analysis in ecology
Data properties? Multivariate (high-dimensional) The question? Abundance (mean-variance) Ordination Hui et al. (2015) Methods Ecol Evol David Warton – Modernising multivariate analysis in ecology
Data properties? Abundance (mean-variance) The question? Test (env impact) Multivariate (high-dimensional) row-resample GLMs Composition Explain spp variation Predict (new sites or spp) Classify Ordination Study correlations David Warton – Modernising multivariate analysis in ecology Latent variable models also see Pledger & Arnold (2014, J Comp Graph Stats)
Data properties? Abundance (mean-variance) Composition Is there a change in community composition? (relative abundance) 64 Hemiptera species Explain spp variation Predict (new sites or spp) Classify Ordination Study correlations David Warton – Modernising multivariate analysis in ecology 36 sites The question? Test (env impact) Multivariate (high-dimensional) 0 0 0 2 0 0 3 0 0 4 0 0 0 ⁞ 4 5 1 2 0 0 0 0 0 3 0 ⁞ 1 0 0 0 0 1 1 ⁞ 0 1 0 0 0 0 1 ⁞ 0 0 1 0 0 0 0 0 12 ⁞ 0 0 0 0 2 6 1 ⁞ 1 0 0 0 1 0 ⁞ 0 0 0 0 3 0 0 0 0 ⁞ 0 0 1 0 0 0 0 ⁞ 0 … … … … 0 1 0 0 2 0 0 0 0 ⁞ 0
Data properties? Abundance (mean-variance) The question? Test (env impact) Composition Explain spp variation Predict “row effect”, ai (new sites or spp) Classify Ordination Study correlations David Warton – Modernising multivariate analysis in ecology Multivariate (high-dimensional) Is there a change in community composition? (relative abundance) 64 Hemiptera species 0 0 0 2 0 0 3 0 0 4 0 0 0 ⁞ 4 5 1 2 0 0 0 0 0 3 0 ⁞ 1 0 0 0 0 1 1 ⁞ 0 1 0 0 0 0 1 ⁞ 0 0 1 0 0 0 0 0 12 ⁞ 0 0 0 0 2 6 1 ⁞ 1 0 0 0 1 0 ⁞ 0 0 0 0 3 0 0 0 0 ⁞ 0 0 1 0 0 0 0 ⁞ 0 … … … … 0 1 0 0 2 0 0 0 0 ⁞ 0
Data properties? Abundance (mean-variance) The question? Test (env impact) Multivariate (high-dimensional) row-resample GLMs Composition Warton et al. (in prep) Explain spp variation Predict (new sites or spp) Classify Ordination Study correlations David Warton – Modernising multivariate analysis in ecology Latent variable models also see Pledger & Arnold (2014, J Comp Graph Stats)
Data properties? The question? Abundance (mean-variance) Composition David Warton – Modernising multivariate analysis in ecology Multivariate (high-dimensional)
Data properties? The question? Abundance (mean-variance) Composition David Warton – Modernising multivariate analysis in ecology Multivariate (high-dimensional) Coral data: effect on total abundance or composition?
Data properties? The question? Abundance (mean-variance) Composition David Warton – Modernising multivariate analysis in ecology Multivariate (high-dimensional) Coral data: effect on total abundance or composition?
Data properties? The question? Abundance (mean-variance) Composition David Warton – Modernising multivariate analysis in ecology Multivariate (high-dimensional) Coral data: effect on total abundance or composition?
Data properties? Abundance (mean-variance) Why do different species respond differently? Composition 64 Hemiptera species Explain spp variation Predict (new sites or spp) Classify Ordination Study correlations David Warton – Modernising multivariate analysis in ecology 36 sites The question? Test (env impact) Multivariate (high-dimensional) 0 0 0 2 0 0 3 0 0 4 0 0 0 ⁞ 4 5 1 2 0 0 0 0 0 3 0 ⁞ 1 0 0 0 0 1 1 ⁞ 0 1 0 0 0 0 1 ⁞ 0 0 1 0 0 0 0 0 12 ⁞ 0 0 0 0 2 6 1 ⁞ 1 0 0 0 1 0 ⁞ 0 0 0 0 3 0 0 0 0 ⁞ 0 0 1 0 0 0 0 ⁞ 0 … … … … 0 1 0 0 2 0 0 0 0 ⁞ 0
Data properties? Abundance (mean-variance) The question? Test (env impact) Composition Multivariate (high-dimensional) Why do different species respond differently? → species traits (zj) as predictors Explain spp variation Predict (new sites or spp) Classify Ordination Study correlations David Warton – Modernising multivariate analysis in ecology climate x trait interactions (bxz) explain differences in species response, “the fourth corner” Brown et al. (2014), Warton et al. (2015) Methods in Ecology and Evolution
Data properties? Abundance (mean-variance) The question? Test (env impact) Composition Explain spp variation Predict (new sites or spp) Classify Ordination Study correlations David Warton – Modernising multivariate analysis in ecology Multivariate (high-dimensional) Why do different species respond differently? → species traits (zj) as predictors
Data properties? Abundance (mean-variance) The question? Test (env impact) Multivariate (high-dimensional) row-resample GLMs Composition Explain spp variation Traits (zj) as predictors, interaction with xi: Predict (new sites or spp) Classify Ordination Study correlations David Warton – Modernising multivariate analysis in ecology Latent variable models also see Pledger & Arnold (2014, J Comp Graph Stats)
Data properties? Abundance (mean-variance) The question? Test (env impact) Multivariate (high-dimensional) row-resample GLMs Composition Explain spp variation Traits (zj) as predictors, interaction with xi: Predict (new sites or spp) LASSO, random effects, neural nets, … (Harris 2015, Methods in Ecology and Evolution ) Classify Ordination Study correlations David Warton – Modernising multivariate analysis in ecology Latent variable models also see Pledger & Arnold (2014, J Comp Graph Stats)
Data properties? Abundance (mean-variance) The question? Test (env impact) Multivariate (high-dimensional) row-resample GLMs Composition Explain spp variation Traits (zj) as predictors, interaction with xi: Predict (new sites or spp) LASSO, random effects, neural nets, … Classify (species by response, or sites by composition) Ordination Latent variable models Study correlations David Warton – Modernising multivariate analysis in ecology (Harris 2015, Methods in Ecology and Evolution ) also see Pledger & Arnold (2014, J Comp Graph Stats)
Data properties? Abundance (mean-variance) The question? Test (env impact) Multivariate (high-dimensional) row-resample GLMs Composition Explain spp variation Traits (zj) as predictors, interaction with xi: Predict (new sites or spp) LASSO, random effects, neural nets, … Classify Finite mixture models Ordination Latent variable models Study correlations David Warton – Modernising multivariate analysis in ecology (Harris 2015, Methods in Ecology and Evolution ) Pledger & Arnold (2014, JCGS), Dunstan et al. (2011 Ecol Model…) also see Pledger & Arnold (2014, J Comp Graph Stats)
Data properties? Abundance (mean-variance) The question? Test (env impact) Multivariate (high-dimensional) row-resample GLMs Composition Explain spp variation Traits (zj) as predictors, interaction with xi: Predict (new sites or spp) LASSO, random effects, neural nets, … Classify Finite mixture models Ordination Latent variable models Study correlations David Warton – Modernising multivariate analysis in ecology (Harris 2015, Methods in Ecology and Evolution ) Pledger & Arnold (2014, JCGS), Dunstan et al. (2011 Ecol Model…) also see Pledger & Arnold (2014, J Comp Graph Stats)
Data properties? Abundance (mean-variance) The question? Test (env impact) Multivariate (high-dimensional) Can use a latent variable model (“biplot”): Composition Explain spp variation Predict (new sites or spp) Classify Ordination Study correlations David Warton – Modernising multivariate analysis in ecology Warton et al. (2015, Trend Ecol Evol)
Data properties? Abundance (mean-variance) GLM+. . . The question? Test (env impact) Or graphical modelling (using a Gaussian copula): Composition Explain spp variation Predict (new sites or spp) Multivariate (high-dimensional) copula S-1 assumed to be sparse: Classify Ordination Study correlations David Warton – Modernising multivariate analysis in ecology
Data properties? Abundance (mean-variance) GLM+. . . The question? Test (env impact) Multivariate (high-dimensional) copula Or graphical modelling (using a Gaussian copula): Composition Explain spp variation Predict (new sites or spp) Classify Ordination S-1 assumed to be sparse, estimate to maximise which can be simplified to an iteratively reweighted GLASSO on the Zij Study correlations David Warton – Modernising multivariate analysis in ecology Popovic & Warton (in review)
Data properties? Abundance (mean-variance) GLM+. . . The question? Test (env impact) Multivariate (high-dimensional) copula row-resample GLMs Composition Explain spp variation Traits (zj) as predictors, interaction with xi: Predict (new sites or spp) LASSO, random effects, neural nets, … Classify Finite mixture models Ordination Latent variable models Study correlations Graphical models (and LVMs) David Warton – Modernising multivariate analysis in ecology (Harris 2015, Methods in Ecology and Evolution ) also see Pledger & Arnold (2014, J Comp Graph Stats)
Data properties? Abundance (mean-variance) GLM+. . . The question? Test (env impact) Multivariate (high-dimensional) GEE, hierarchical, copula row-resample GLMs Composition Explain spp variation Traits (zj) as predictors, interaction with xi: Predict (new sites or spp) LASSO, random effects, neural nets, … Classify Finite mixture models Ordination Latent variable models Study correlations Graphical models (and LVMs) David Warton – Modernising multivariate analysis in ecology (Harris 2015, Methods in Ecology and Evolution ) also see Pledger & Arnold (2014, J Comp Graph Stats)
Software for multivariate models Test (env impact) Composition resample GLMs (anova. manyglm on mvabund) (manyany on mvabund) Explain spp variation Traits (zj) as predictors, interaction with xi: (traitglm on mvabund) Predict (new sites or spp) LASSO (glm 1 path on mvabund, or glmnet), random effects (lme 4 etc), neural nets (mistnet) Classify Finite mixture models (speciesmix, RCPmod) Ordination Latent variable models (boral, HMSC) Study correlations Graphical models (and LVMs) David Warton – Modernising multivariate analysis in ecology
The case of the missing model: The modernisation of multivariate analysis in ecology • Introduction • The Bray-Curtis distance and other 1980’s memorabilia • Building a model for multivariate abundances in ecology • A 2020 vision for multivariate analysis in ecology David Warton – Modernising multivariate analysis in ecology
A 2020 vision for multivariate analysis Less of this: More of this: David Warton – Modernising multivariate analysis in ecology
To-do list for 2020: Abundance (mean-variance) Test (env impact) Composition Explain spp variation Predict (new sites or spp) Classify Ordination Study correlations David Warton – Modernising multivariate analysis in ecology Multivariate (high-dimensional)
To-do list for 2020: Abundance (mean-variance) Multivariate (high-dimensional) Test (env impact) Composition Explain spp variation Predict (new sites or spp) Classify Ordination Study correlations David Warton – Modernising multivariate analysis in ecology • Model-based inference with many parameters
To-do list for 2020: Abundance (mean-variance) Multivariate (high-dimensional) Test (env impact) Composition Explain spp variation Predict (new sites or spp) Classify Ordination Study correlations David Warton – Modernising multivariate analysis in ecology • Model-based inference with many parameters • Classification + high-dimensional response
To-do list for 2020: Abundance (mean-variance) Multivariate (high-dimensional) Test (env impact) Composition Explain spp variation Predict (new sites or spp) Classify Ordination Study correlations David Warton – Modernising multivariate analysis in ecology • Model-based inference with many parameters • Classification + high-dimensional response • Copula inference for high-dimensional counts
To-do list for 2020: Abundance (mean-variance) Multivariate (high-dimensional) Test (env impact) Composition • Model-based inference with many parameters Explain spp variation • Classification + high-dimensional response Predict (new sites or spp) • Hierarchical classification using models Classify Ordination Study correlations David Warton – Modernising multivariate analysis in ecology • Copula inference for high-dimensional counts
To-do list for 2020: Abundance (mean-variance) Multivariate (high-dimensional) Test (env impact) Composition • Model-based inference with many parameters Explain spp variation • Classification + high-dimensional response Predict (new sites or spp) • Hierarchical classification using models Classify Ordination Study correlations David Warton – Modernising multivariate analysis in ecology • Copula inference for high-dimensional counts • Structured correlation across sites (spatial etc)
To-do list for 2020: Abundance (mean-variance) Multivariate (high-dimensional) Test (env impact) Composition • Model-based inference with many parameters Explain spp variation • Classification + high-dimensional response Predict (new sites or spp) • Hierarchical classification using models Classify Ordination Study correlations David Warton – Modernising multivariate analysis in ecology • Copula inference for high-dimensional counts • Structured correlation across sites (spatial etc) • Spread the word to ecologists…
Spreading the word… • • • Free well-documented software (mvabund etc) Youtube video → Blog posts Textbook (to write in 2016) Workshops Eco-Stats conferences David Warton – Modernising multivariate analysis in ecology
Acknowledgements • UNSW Eco-Stats Group and Collaborators • Australian Research Council – Future Fellowship scheme (FT 120100501) – Discovery scheme (DP 0985886, DP 120100882, DP 150100823) • UNSW – School of Mathematics and Statistics – Evolution and Ecology Research Centre • Ecologists who shared their data! David Warton – Modernising multivariate analysis in ecology
- Oracle forms modernisation
- Role of rrrlf in promotion and development of libraries
- Modernization and globalization
- Modernisation theory
- Modernization theory
- Best worst and average case
- Ms missing model
- Long case vs short case
- Binary search time complexity average case
- Glennan building cwru
- Bubble sort best case and worst case
- It project failure case study
- Bubble sort best case and worst case
- Bubble sort best case and worst case
- What is ambiguous case for law of sines
- Concentric sector and multiple nuclei model
- Hát kết hợp bộ gõ cơ thể
- Ng-html
- Bổ thể
- Tỉ lệ cơ thể trẻ em
- Voi kéo gỗ như thế nào
- Tư thế worm breton là gì
- Chúa yêu trần thế
- Các môn thể thao bắt đầu bằng tiếng nhảy
- Thế nào là hệ số cao nhất
- Các châu lục và đại dương trên thế giới
- Công thức tính thế năng
- Trời xanh đây là của chúng ta thể thơ
- Mật thư tọa độ 5x5
- 101012 bằng
- Phản ứng thế ankan
- Các châu lục và đại dương trên thế giới
- Thơ thất ngôn tứ tuyệt đường luật
- Quá trình desamine hóa có thể tạo ra
- Một số thể thơ truyền thống
- Cái miệng nó xinh thế chỉ nói điều hay thôi
- Vẽ hình chiếu vuông góc của vật thể sau
- Biện pháp chống mỏi cơ
- đặc điểm cơ thể của người tối cổ
- Thứ tự các dấu thăng giáng ở hóa biểu
- Vẽ hình chiếu đứng bằng cạnh của vật thể
- Vẽ hình chiếu vuông góc của vật thể sau
- Thẻ vin
- đại từ thay thế
- điện thế nghỉ
- Tư thế ngồi viết
- Diễn thế sinh thái là
- Dot
- Số.nguyên tố
- Tư thế ngồi viết
- Lời thề hippocrates
- Thiếu nhi thế giới liên hoan
- ưu thế lai là gì
- Hổ sinh sản vào mùa nào
- Khi nào hổ con có thể sống độc lập
- Sơ đồ cơ thể người
- Từ ngữ thể hiện lòng nhân hậu
- Thế nào là mạng điện lắp đặt kiểu nổi
- Use case model
- Use case narratives
- Use case specification example
- Neuman system model
- Private equity case study presentation
- What is the butler model
- Component diagram of hospital management system
- Er model for banking system
- Daniya champagne
- To determine missing z values for a statistical surface
- Reasons for angles
- Discrete missing values spss